Softcore Processor Optimization According to Real-Application Requirements

Nowadays, embedded processor cores are integrated into most system-on-chip (SoC). Processor cores can be designed to be dedicated for an SoC. However, reusing of generic processors is often preferred due to time to market constraint. Such processors have drawbacks in terms of hardware complexity and power consumption. Indeed, some of their instructions and hardware resources are useless. These area and energy inefficiencies are problematic for low-cost and low-energy systems. In this paper, we propose a methodology for automatically reducing processor functionalities and the resulting hardware complexity according to real-application requirements. This approach was evaluated on two open-source processor cores. The results show that the average area and power consumption savings are over 20% on both application-specific integrated circuit (ASIC) and field-programmable gate array (FPGA) technologies.